* martin f krafft <[EMAIL PROTECTED]> [071101 23:45]: > also sprach Micah Anderson <[EMAIL PROTECTED]> [2007.05.20.0110 +0200]: > > Have you looked at what would break if this test was removed? I'm > > not convinced that its unnecessary, and removing it will only > > solve problems and not create new ones. However, if you've > > determined that this is so, I'd certainly love to hear that. > > Well, when the server side rdiff does not match the local version, > it bitches: > > piper:~> rdiff-backup --include /etc --exclude '/*' / [EMAIL > PROTECTED]::/srv/backups/piper > Warning: Local version 1.1.9 does not match remote version 1.1.5. > > it then works. My assumption is simply that since rdiff-backup > itself does a version comparison and should know better which > versions are compatible with each other, adding a 1:1 requirement to > backupninja may be more limiting than necessary.
We did look at removing the minor version check for rdiff-backup because it was reported that upstream was going to attempt to maintain archive version compatability between minor versions. So we changed things so it would allow you to backup between minor versions, such as 1.1.5 and 1.1.12. I then ran some tests and the error that was created was cryptic to say the least: "Exception 'too many values to unpack' raised of class". I did some research and the rdiff-backup authors said[1] that this error is a result of "not using the same version of rdiff-backup on both sides" since "an important change was made to the network protocol in 1.1.12 to fix a long-standing bug. If you use 1.1.12 on both sides, then you will not have this problem." Hmm, ok thats not nice. Since a lot of backupninja users don't really know that having the same version of rdiff-backup on both sides is generally a good idea, and I didn't see a warning about this, just an ugly error I am afraid that if we allow this to happen, we will end up getting a lot of reports of failures due to version mismatches. > In this very instance, it disables me from tightening the > password-less login for the backup user via authorized_keys. I suppose that this version check could be overridden as an option in the config file. It makes me worry a little about feature creep, especially at a time when there is little resources available to add new features. Micah 1. http://lists.gnu.org/archive/html/rdiff-backup-users/2007-07/msg00091.html
signature.asc
Description: Digital signature
